I've discussed this till I've gone blue, but there are also other glaring errors in Coyle's paper that have a marked bearing on his conclusions. Even Jackhammer, the staunchest of LA fans, admitted that a high school maths teacher could see Coyle was wrong.
This is Table 2 from Coyle's paper summarizing his data:
Coyle's conclusions were basically Improved Efficiecy + Weight Loss = 18% Improvement in Power-to-Weight Ratio.
As Kreb's Cycle and Eva Maria have pointed out, the efficiency calculations are highly suspect.
As you can see from the Table 2 data, percentage body fat was not measured in 1999 and lean and gross body weights are actually heavier than either 1992 or 1993. So the weight loss part of the equation is clearly incorrect. (Coyle based his weight loss argument on Armstrong's estimated racing weight of 72-74 kg, which was neither measured and nor was it relevant considering all data is from preseason and not racing season).
As you can also see from Table 2, weight and power are listed which gives power-to-weight ratios of 4.74 in preseason 1992, 4.99 in preseason 1993, and 5.07 in preseason 1999. This results in improvements of 6.9% compared to 1992 preseason and 1.6% compared to preseason 1993 - nowhere near the reported 18%. Again, Coyle comes up with this figure because for 1999 he used the measured preseason power with Armstrong's estimated racing weight rather than the measured preseason body weight.
So when you look at Coyle's conclusions, efficiency is highly questionable, weight loss is incorrect and the improvement in power-to-weight ratio is also incorrect. A fine bit of rigorous scientific study IMO!